TB pixel clustering - lower learning rate, 10 steps, 3 replicates
Candace Liu
January 14, 2021
- Gaussian blur sigma=2, k=20 (pixel metaclusters)
- z-score cap=3 for hierarchical clustering
- For each pixel, divide each marker by total (each pixel sums to 1)
- learning rate: 0.01 to 0.001
Rep 1
Histogram of counts in pixel clusters
Heatmap of pixel clusters x markers
Heatmap of pixel metaclusters x markers
Breakdown of clusters per point
Rep 2
Histogram of counts in pixel clusters
Heatmap of pixel clusters x markers
Heatmap of pixel metaclusters x markers
Breakdown of clusters per point
Rep 3
Histogram of counts in pixel clusters
Heatmap of pixel clusters x markers
Heatmap of pixel metaclusters x markers
Breakdown of clusters per point
